In the global DRAM market in the second quarter this year, Samsung Electronics kept the No. 1 spot, while Micron was found to have closed in on SK hynix by 1 percentage point.
According to Counterpoint Research's "Global Memory Tracker" on the 4th, Samsung Electronics maintained the lead with a 39% market share by DRAM revenue in the second quarter this year. The rise in commodity DRAM prices and expansion of the high bandwidth memory (HBM) business were analyzed to have supported the results.
SK hynix's revenue grew 214% from a year earlier, but its market share fell to 26% from 39% in the same period last year.
Micron, in third place, saw revenue increase to about five times from a year earlier and posted a 25% share, narrowing the gap with SK hynix to 1 percentage point.
Counterpoint Research cited an HBM-centric product portfolio as the reason behind SK hynix's share decline. The firm said its HBM bit shipments and revenue mix are higher than rivals, making it more affected by the price drop of existing HBM3E, and that long-term supply agreements (LTA) signed before the memory price upswing also limited the increase in average selling price (ASP).
While commodity DRAM prices rose sharply from the previous quarter in the second quarter this year, HBM prices were found to be lower than last year as the HBM3E price decline overlapped with a delay in the HBM4 launch.
However, Counterpoint Research projected that once HBM4 shipments ramp up starting in the third quarter this year, the HBM price decline will slow due to an improved product mix, and next year HBM prices will rise again in line with the upward trend in commodity DRAM prices.
China's CXMT ranked fourth with a 7% share, recording 716% year-over-year revenue growth on the back of expanding domestic demand centered on commodity DRAM and capacity additions.
Counterpoint Research said, "If CXMT expands its production capacity to a level that can meet demand from major global PC makers and other large customers, its market share a year from now could look very different," adding, "The key is not whether CXMT breaks into the 'big three' of memory, but when."